Clive McDowell schrieb:

I'll jump in here since I asked a similar question yesterday (pay attention to 
the list Ronan!). The SA check is the last in the
list of acl data checks. All helo sanity, rbls, virus checks etc. take place 
first. We seem to have a strange problem with our
SA server between midnight and ~3.00am when it appears to stop responding to 
connection requests. We haven't got to the bottom
of what is happening on the SA server but on a general point (in our situation) 
it would be best to defer messages if the SA
server is down rather than passing them unchecked. Our current acl is -

  warn    message       = X-Spam-Score: $spam_bar ($spam_score)\n\
                          X-Spam-Score-Int: $spam_score_int
          condition     = ${if <{$message_size}{80k}{1}{0}}
          spam          = nobody:true

What do we have to change to defer a message if we cannot pick up a spam score?

Hi Clive,

you have to implement a tri-state logic, so you can differentiate between ham, spam and a message that wasn't scanned due to SA outages.

You can achieve this with something like this:

  # Scan message content for spam characteristics. We only scan mails
  # that are not larger than MESSAGE_SIZE_SPAM_MAX. Defer if the content
  # scanner is not running.
  warn
    condition   = ${if > {$message_size}{MESSAGE_SIZE_SPAM_MAX} \
                      {true}{false}}
    set acl_m2  = ee_not_spam_scanned
    set acl_m9  = 0
  warn
    condition   = ${if > {$message_size}{MESSAGE_SIZE_SPAM_MAX} \
                      {false}{true}}
    set acl_m9  = 1
    spam        = nobody:true
    set acl_m9  = 0
    set acl_m3  = ${eval:$acl_m3 + $spam_score_int}
  defer
    condition   = $acl_m9

  # Deny if the spam score reaches the spam threshold.
  deny
    condition   = ${if >= {$acl_m3}{SPAM_THRESHOLD} \
                      {true}{false}}
    set acl_m2  = ee_threshold_data


About the behaviour you see: Ask yourself, what happens in this timeframe, when connections to SA timeout. Probably SA is busy, as I don't expect that you get more spam in this timeframe than in any other. For example, are you training your bayes database in this timeframe? Questions like this one should put you on the right track.


HTH,
Patrick Eisenacher

--
## List details at http://www.exim.org/mailman/listinfo/exim-users ## Exim details at http://www.exim.org/
## Please use the Wiki with this list - http://www.exim.org/eximwiki/

Reply via email to